Articles by Grizzly Peak Software

Five-layer guardrail architecture for AI document extraction pipelines that fabricate data instead of failing visibly.

Build a local-first MCP documentation server so your AI coding assistant actually knows your codebase, APIs, and team conventions.

REST APIs weren't built for AI agents. MCP servers are the semantic layer that lets AI discover and invoke your tools. Here's why adoption is inevitable.

After 40+ years in software, here are the engineering skills AI has devalued and the ones worth doubling down on right now.

Five products, no employees, under $1K/month. Here's the exact stack, AI tooling, and philosophy behind running it all solo.

80+ books, 70 audiobooks, one AI-assisted pipeline. Here's why the "AI content is slop" argument confuses the tool with the operator.

Your AI writes fast. It also writes JWT bugs, hardcoded secrets, and SQL injection. Here's what the scanners actually found.

AI delivery pipelines need governance for accountability and compliance. Add traceability, approval gates, and audit trails without crushing velocity.

I built an AI chatbot that diagnoses car problems and generates repair content. Here's what I learned about trust, accuracy, and domain-specific AI.

A Node.js engineer explains why single-agent AI systems fail at scale and builds a multi-agent orchestrator with smart routing to fix it.

A hands-on walkthrough of fine-tuning Mistral-7B on personal blog posts using LoRA on an NVIDIA DGX Spark — from training data to local deployment, in one evening.

Low-code platforms promised to democratize software development. AI-augmented engineering just made that promise obsolete — and cheaper.

LangExtract brings schema enforcement, source grounding, and multi-pass extraction to any LLM. Here's how to build a real PDF intelligence pipeline.

I built an open-source EPUB reader in one night using Claude Code — drag, drop, read, and analyze your book before it goes live.

Frameworks hide the loop. This tutorial builds a working Python AI agent in 50 lines, then adds a tool registry, memory, and multi-agent patterns.

Last month I was looking at my analytics for AutoDetective.ai and noticed something I wasn't expecting. A non-trivial chunk of referral traffic was coming...

I built a job board because I kept getting the same email.

AI Overviews crushed my reference content. Experience-based articles held. I rebuilt my SEO strategy around what LLMs can't replicate.

Most developers build one product at a time and market it in isolation. I know because I did this for years and it's an incredibly inefficient way to...

I've spent more on AI tools than groceries in the last 18 months. In Alaska, that's a lot of milk. Here's what actually earned its keep — and what was a waste.